Zachary Tam,Karthik Dharmarajan,Tianshuang Qiu,Yahav Avigal,Jeffrey Ichnowski,Ken Goldberg
Abstract:In logistics, the ability to quickly compute and execute pick-and-place motions from bins is critical to increasing productivity. We present Bin-Optimized Motion Planning (BOMP), a motion planning framework that plans arm motions for a six-axis industrial robot with a long-nosed suction tool to remove boxes from deep bins. BOMP considers robot arm kinematics, actuation limits, the dimensions of a grasped box, and a varying height map of a bin environment to rapidly generate time-optimized, jerk-limited, and collision-free trajectories. The optimization is warm-started using a deep neural network trained offline in simulation with 25,000 scenes and corresponding trajectories. Experiments with 96 simulated and 15 physical environments suggest that BOMP generates collision-free trajectories that are up to 58 % faster than baseline sampling-based planners and up to 36 % faster than an industry-standard Up-Over-Down algorithm, which has an extremely low 15 % success rate in this context. BOMP also generates jerk-limited trajectories while baselines do not. Website: <a class="link-external link-https" href="https://sites.google.com/berkeley.edu/bomp" rel="external noopener nofollow">this https URL</a>.
What problem does this paper attempt to address?
This paper attempts to solve the problem of robots quickly calculating and performing pick - and - place actions from deep bins in logistics scenarios. Specifically, the authors propose Bin - Optimized Motion Planning (BOMP), a motion planning framework aimed at equipping six - axis industrial robots with long - nosed suction cup tools to remove boxes from deep bins. BOMP takes into account the robot's kinematics, drive limitations, the size of the grasped box, and the changing in - bin height map to quickly generate time - optimized, acceleration - limited, and collision - free trajectories.
### Problem Background
In logistics scenarios, thousands of parcels need to be processed every day. Reducing the processing time for each parcel can significantly improve the productivity of robots. Especially in deep - bin environments, robots are usually equipped with long - nosed suction cup tools (such as the "bluction" tool proposed by Huang et al.) so that they can reach and operate items inside the bin. In addition, items inside the bin will move after each grasp, so a strategy is needed to quickly calculate new pick - and - place paths using the latest sensor data (such as RGBD camera images).
### Existing Methods and Their Limitations
Existing methods include:
1. **Heuristic Planning**: For example, the Up - Over - Down algorithm. Although it is simple to implement and has a short calculation time, vertically lifting the box in a deep - bin environment may fail due to kinematic infeasibility.
2. **Optimization - Based Motion Planning**: Find the best or fastest collision - free trajectory by solving an optimization problem, but the calculation time is long.
3. **Sampling - Based Motion Planning**: Randomly sample collision - free waypoints to find a path. Although the success rate is relatively high, the calculation time is also long.
### Main Contributions of BOMP
To address the above challenges, BOMP makes the following contributions:
1. **Proposes a time - optimized, acceleration - limited motion planning algorithm** suitable for picking up boxes from deep bins. Experiments show that the trajectories generated by BOMP are 36% faster than the Up - Over - Down benchmark and 58% faster than MPT (sampling - based motion planning).
2. **Develops an end - to - end in - bin pick - up pipeline** to iteratively remove boxes using BOMP.
3. **Trains a deep neural network** to predict the initial trajectory based on the environmental height map, grasping box parameters, number of trajectory segments, and trajectory end points, in order to accelerate the convergence of the optimizer.
4. **Provides experimental data from 96 simulation environments and 15 physical environments** to verify the effectiveness of BOMP.
### Method Overview
The core of BOMP is to combine optimization - based motion planning and deep - learning warm - start techniques. The specific steps are as follows:
1. **Target Box Selection and Shape Estimation**: Detect the box at the highest point through RGBD images, and use the SAM model to generate a segmentation mask, and then estimate the size of the box.
2. **Suction Cup Grasping Point Selection**: Select the most appropriate suction cup grasping point and ensure that the robot will not collide with the environment.
3. **Optimization Formulation**: Solve the optimal trajectory through a nonlinear optimization problem, taking into account the robot's kinematics, dynamic limitations, and obstacle avoidance.
4. **Collision Detection**: Accelerate the collision detection process and only check for collisions between the suction cup tool, the grasped box, and the environment.
5. **Deep - Learning Warm - Start**: Use the trained neural network to predict the initial trajectory and segmented time steps to accelerate the convergence of the optimizer.
### Experimental Results
The experimental results show that BOMP outperforms existing benchmark methods in both calculation time and execution time, while successfully generating collision - free trajectories and improving the overall success rate.
Through these improvements, BOMP can efficiently complete pick - and - place tasks in complex deep - bin environments, significantly improving the productivity of robots in logistics scenarios.